Share your A.I Text to image "creations"

[f2j]

Hey there
So I've been playing with text to image A.I lately, and especially with MidJourney.
I though it could be cool if we share here what the A.I generated and the prompt used for that purpose.

If you haven't tried yet, you can go to https://www.midjourney.com/ , you'll need a Discord account to log in. There is a free trial for generating 200 images.
When you log in the website with your discord account, it will join you to the discord server, where you can send your requests to the A.I by writing "/imagine the description of what you want the AI to create", in an appropriate channel (starting with newbies-##, or in DM with the MidJourney Bot).

The bot will render 4 images, in one minute.
Then you can choose to ask variations on one or severall of the proposed images.
Or you can also upscale one or severall images.


So here is a first example :
Prompt : a beautifull cannabis crop. The plants are in pots. Over the plants is LED lightning for plant growing. 8K, HD. Sharpness, Octane render
(the A.I did not respect a part of my prompt but whatever ^^)

Prompt : a dream landscape in asiatic style

Your turn now !

 

[f2j]

Prompt : dry herb vaporizer. HD, 8K, Cinematic. Sharpness. Octane render

Prompt : a scary creature with a lot of arms, a long and spiky tail. In the style of [H.R Giger]. Octane render. Sharpness. HD. 8K.

 

[Abele Rizieri Ferrari]

marijuana goddess Mother Earth weed

alien stoned high weed space

zombie anatomy

ancient tribe smoking weed

 

[f2j]

Nice Marijuana goddess

Prompt : [Emmanuel Macron] as a human squirrel

Prompt : a bunny from the future

 

[f2j]

Prompt : a dream land with a castle in the mountain in the background, girl in the foreground, fantasy, concept art, night, starrly sky --hd --ar 21:30

Prompt : a japanese Samourai contemplating a modern looking japanese city in the background. He wears a sword on his back. He wears a samourai mask.

 

[passenger]

with midjourney...

 

[Siebter]

Medieval french church on a winter day at the edge of a forest late Robert Crumb style colored

Aztek queen bee drinking beer with an onyx elephant rough paintbrush style

More to come soon. :-)

 

[Siebter]

Underwater dance club in the 1920s, many people dancing, many lush plants, some smoke, golden light, red light, yellow light, amber light, very detailed

 

[f2j]

Prompt : a little girl, wathcing herself into a mirror. Her reflection is her face as an old lady.

 

[Abele Rizieri Ferrari]

person sitting behind a computer. A tube runs from their mouth. The room is smokey made this one in Dall-E

This is what mid journey came up with:

Some I did in mid journey:

colored frog sitting on a lily in a tropical pond rainforest mood lighting lush vegetation green gold

futuristic utopian city with flying car and lots of plants

rainbow unicorn translucent

 

[Shit Snacks]

I finally tried a few at midjourney, (I hate discord so much, probably because I'm not using the app, but trying to go to the website on my phone browser lol) these were some of the best I think...

Mermaid Squid Pirate Goddess:

Pirate Ships around Archipelago with Deep Seas and Swirling Skies:

 

[Abele Rizieri Ferrari]

back of hooded monk seen in overgrown forest. Cloak. Mood lighting. Realistic

Andy Warhol can of worms

steampunk ufo flying saucer

Alf sitting at a bar drinking a beer. Ultra-realistic

ancient warrior demon mask. Metallic. Dmt vision. Detailed. Geometric pattern. Colorful.

expressionist painting of the artist's studio

watercolor painting of an orange sunset over a blue harbor with some sailing ships

 

[passenger]

If this develops at this speed and extends to video footage, good night sweet truths.

 

[Siebter]

„Grocery shopping“

 

[Abele Rizieri Ferrari]

angel and devil getting high smoking weed

two robots drinking a cup of coffee at a coffeeshop

(Omg aren't they cute )

Human tree

Spoiler: CW: kind of dark

 

[Siebter]

„Discussing climate change“

---

Very nice tool: → Prompter – it lets you put various parameters into a google doc that will turn them into a finished prompt.

 

[Siebter]

Guys, what's happening here? Don't stop sharing, okay? :-)

I must admit I fell pretty deep into the midjourney rabbit hole, constantly throwing ideas at this weird and wonderful machine... I'm struggling a bit with focusing on one image, instead I create entire series of images. I'm also still learning how to create proper prompts and how to steer the A.I. in a specific direction.

Anyway, here are a few excerpts of the stuff I've been coming up with over the last few days.

From my „Religion“ series:

„Bishop in chair III“

„Bishop in chair IV“

„Religion machines - Christianity I“

„The Parousia - I - Annunciation b“

„II - Rebuilding IV“

The „Bathtub“ series:

„Ib - Testing the waters“

„Vb - It's bath time“

Animal factories series:

„Bird factory I“

„Bird factory II - abandoned“

„Panda bear factory“

Chem lab bee hive series:

„II“

„Detail II“

 

[f2j]

I have a hard time orienting the IA towards what i'm picturing. But that is also what's awesome with these AI, there's always something cool that you did not think of, some form of random creativity. And also I love the AI ability to create ambiances

Did you use a promt generator or something ?

 

[Siebter]

Did you use a promt generator or something ?

Not really, at least not to create an entire prompt, but I find them useful to look up what kind of options I even have. The various ways to set the light, the camera lens or choose an artistic style are pretty much endless, so having some kind of a lookbook to browse through makes things a bit easier.

There are some key things that are important to keep in mind, for example the layout of a good prompt, which should be

[URL] -- scenery -- arguments

Some overlook the „--stylize“ argument, lowering its value is especially helpful to get closer to what you have in mind. The default value is 2500, setting it to 625 turns it off (as off as it can be under these circumstances).

--stylize 625 stay close to prompt scenery --stylize 5000 be more playful --stylize 60000 be artistic and experimental possible values: 625, 1250, 2500, 5000, 10000, 20000, 40000, 60000

the „--chaos“ parameter is a bit different, it determines how much artistic randomness is mixed in (not sure what the default is).

--chaos (0, 10, 25, 50, 75, 100)

Weighting elements is very useful too, though not always as precise as one would like to hope:

cat::6 dog::3 make cats twice as important as dogs

Note that this is not a quantifying value, it sets how *important* an element should be.

You can also try to use brackets to set elements apart from each other:

[a cat in a tree] + [skyline background, traffic lights]

...and combine them with weight parameters too:

[a cat in a tree]::10 + [skyline background, traffic lights]::5

How each element is sequenced within a prompt makes a huge difference. You can create an image in the style of Jackson Pollock or create an image that is 3D, but has objects that are painted by Jackson Pollock (see my „Bathtub“ series). Elements at the end of a prompt are less important than those at the beginning (which is where weighting comes into play).

One should be mindful when coupling elements with adjectives etc.:

antique brass candlestick different from → antique-brass-candlestick ornate massive shadowed sentient different from → ornate, massive, shadowed, sentient

It's also important to mind the space an image should create, if we don't, things become hard to handle. Is it an open field, a little chamber, a cathedral or a hall? Same goes for the aspect ratio: --ar 16:9, --ar 2:1 or --ar 1:2 will create very different images.

And lastly, keeping things out of an image via the „--no“ parameter can be useful:

--no plants remove / avoid plants

Some cool links:

→ Phraser -- yet another prompt creator, but more intuitive than Prompter
→ An advanced guide to writing prompts for Midjourney (text-to-image)
→ Exploring Midjourney’s AI art style using circles
→ Livia Theodora's MJ references
→ Will Wulfken's MJ styles and keyboard reference

Playing with effects and parameters is a lot of fun, but I always try to focus on the idea I originally had. I do agree though:

But that is also what's awesome with these AI, there's always something cool that you did not think of, some form of random creativity.

Being creative with an A.I. is much more of a dialogue than anything else, so being open minded about suggestions is a bit of a must (and fun!). It's not only the A.I. that is permanently learning with each prompt we enter, learning from Midjourney's output will help us to explain our ideas too.

 

[Abele Rizieri Ferrari]

Meta made a text to video AI: https://makeavideo.studio/

Only some demos now but looks impressive, can also sign up to join.

 

[Siebter]

Meta made a text to video AI: https://makeavideo.studio/

Honestly, that doesn't even seem half as intriguing as Midhourney... It is impressive from a programming standpoint for sure, but the output I find to be very uninspiring.

---

„Bee culture“

„I”

„IV“

„VIIb“


„Last minute in the lounge“

„III“

„IVb”

„V“

„Bishop portraits I“

„Siebter X Banksy collab“

„The Parousia – The End – IV“

...and some untitled stuff:

 

[Abele Rizieri Ferrari]

Honestly, that doesn't even seem half as intriguing as Midhourney... It is impressive from a programming standpoint for sure, but the output I find to be very uninspiring.

Gotta start somewhere. During the period i used mind journey the ai developed so quickly, and from what I can tell they had also spent more hours teaching the ai and getting to know it (in part through the beta testers meta is reaching out to). Also video is of course a different medium than an image so to make it intriguing and create inspiring output it needs a different sort of creativity perhaps. Anyway don't wanna dwell off topic, just thought it was cool and semi related.

 

[florduh]

Lo Fi Nuclear War to Relax and Study to

 

[Abele Rizieri Ferrari]

Lo Fi Nuclear War to Relax and Study to

I can relate to this image

 

[florduh]

Orson Welles as John Wick